Section: Partnerships and Cooperations

International Research Visitors

Visits of International Scientists

• Andrew Barto, Reinforcement learning and intrinsic motivation, University of Massachusetts Amherst, USA (oct 2012)

• Adam White, Reinforcement Learning and Artificial Intelligent group, Computing Science department of the University of Alberta, Canada (September 2012)

• Joseph Modayil, Reinforcement Learning and Artificial Intelligent group, Computing Science department of the University of Alberta, Canada (September 2012)

• Akihiko Yamaguchi, Robotics Lab of Prof. Ogasawara at NAIST in Japan (march 2012)

• Todd Hester, RL and Robotics Lab, Univ. Texas, US (may, june, july 2012)

• Louis ten Bosh, Speech processing, Univ. Radboud, The Netherlands (june 2012)

• Robert Saunders, Design Lab, Faculty of Architecture, University of Sydney, Australia (september 2012)

• Adrien Baranes, Columbia University, NY, USA (october 2012)

• Joshka Boedecker, Asada Lab, Osaka University, Japan (october 2012)

• Olivier Georgeon, Univ. Lyon, France (november 2012)

Internships

• Gennaro Raiola, MSc. Student from Università degli Studi di Napoli Federico II. Parameterized skills are able to map parameters of the task (for instance the 2D position of an object on a table) to the appropriate parameters of a policy for achieving this task. In this project, we use imitation learning to train a Dynamic Movement Primitive (DMP) with several observed trajectories. To achieve generalization, the basis functions in the DMP are expanded so that they span the space of the task relevant parameters. The resulting algorithm is applied to human reaching data, and to generalizing skills on the Nao robot.

• Laura Vogelaar, visiting student from GeorgiaTech and Carnegie Mellon University. Within a stochastic optimization context, we use clustering algorithms to determine features that are relevant to minimizing the cost of executing a skill. Our objective is to enable a robot to autonomously expand its libraries of skills, whilst simultaneously learning which skills can be successfully executed in which contexts.

Visits to International Teams

• Manuel Lopes (December 2012), Willow Garage, Palo Alto, USA: visit to Maya Cakmak to discuss tutoring systems and human-robot internaction.

• Manuel Lopes (December 2012), Bosch Research, Palo Alto, USA: visit to Dejan Pangercic to discuss active learning and human-robot interaction.

• Manuel Lopes (December 2012), Berkely University, USA: visit to Pieter Abbeel to discuss safe exploration methods and inverse reinforcement learning.

• Manuel Lopes (December 2012),

• Clément Moulin-Frier (November 2012), UC Merced, USA: visit to Anne Warlaumont's lab at UC Merced, to discuss about the role and the computational modeling of infraphonology in infant language development. The aim is to initiate a collaboration with Anne Warlaumont and D. Kimbrough Oller (University of Memphis, USA) to computationally study the possible role of intrinsic motivations in infraphonological exploration.

• Olivier Mangin (17/10/2012), Instituto Superior Técnico, Lisbone, Portugal

• Thomas Degris (June 2012), Reinforcement Learning and Artificial Intelligent group, Computing Science department of the University of Alberta, Canada (June 2012)

Participation to Summer/Winter School

• Jonathan Grizou participated to e'NTERFACE 2012, July, 2nd - July, 27th 2012, SUPELEC, Metz, France The 8th International Summer Workshop on Multimodal Interfaces took place on the SUPELEC campus of Metz, France. This one month summer school brought together more than 70 students and experts to work together and foster the development of tomorrow's multimodal research community. Jonathan Grizou enrolled in the Project P1 : "Speech, gaze and gesturing – multimodal conversational interaction with Nao robot", supervised by Graham Wilcock and Kristiina Jokinen (University of Helsinki). This summer school lead to a join publication by the members of the project P1 at the CogInfoCom 2012 conference [34] .

• Jonathan Grizou and Fabien Bénureau participated to the IM-CLeVeR/FIAS Winter School on "Intrinsic Motivation: From Brains to Robots", December 3-8, 2012, Frankfurt Institute for Advanced Studies, Frankfurt am Main, Germany. The school brought together 25 students in the field of intrinsic motivation as well as leaders in the field (among which, Andrew Barto, Minoru Asada, Peter Redgrave, Giorgio Metta and others). Students' time was divided between keynotes in the morning and project work in the afternoon, supervised by the speakers and the school organizers. The school was an opportunity to meet and discuss with researchers and PhD students. It also allowed us to explain and disseminate our work; Pierre-Yves Oudeyer, notably, was an invited speaker. Jonathan Grizou took part in the project "Intrinsic Motivation in Active Perception" while Fabien Benureau participated in "Playful Acquisition of Basic Behavioral skills Machine". The results of the school are highly positive, and some scientific collaborations may directly stem from this event in the future.